ai-virtuoso

Plugin ai-virtuoso from marketplace code-virtuoso in krzysztofsurdy/code-virtuoso. Commands read from the repo's manifest — 16 days ago.

Skills (43)

• llm-evals

  Evaluate and regression-test LLM features and agents with trace error analysis, binary graders, code assertions, validated judges, and CI thresholds. Use when building an eval suite, checking a prompt or model change, or measuring agent quality.

• llm-observability

  Instrument LLM apps and agents with OpenTelemetry GenAI traces, token and cost attribution, sampling, and PII-safe capture. Use when adding tracing, debugging agent runs, or cutting LLM cost and latency.

• godot-upgrade

  Godot 4 minor-version upgrade playbook for hops from 4.0 to 4.7, with behavior changes per version and a verify step. Use to upgrade a Godot project, open it in a newer editor, check changed defaults, or fix glow, physics, .uid or scene-diff surprises after an upgrade.

• godot

  Godot 4 work in a project with project.godot, .gd or .tscn files. Use to verify a change in the running game, fix a headless GDScript test run that hangs or exits 0, hand-edit a .tscn, commit .uid files, or catch an invented API.

• composer-dependencies

  Composer playbook for safe package updates. Use to update dependencies, audit for vulnerabilities, manage composer.lock, set up Dependabot or Renovate, replace abandoned packages, or resolve version conflicts.

• php-upgrade

  Step-by-step PHP upgrade playbook for hops from 7.4 to 8.5, with known traps per version, Rector rule sets, and a CI matrix. Use to upgrade PHP, fix deprecations, check compatibility, or plan a PHP migration.

• symfony-upgrade

  Symfony version upgrade guide using the deprecation-first approach. Use to upgrade to a new minor or major version, fix deprecations, update recipes, check bundle compatibility, or move between LTS versions.

• vercel-ai-sdk

  Decision guide for the Vercel AI SDK (`ai` and `@ai-sdk/*`) in TypeScript, covering the version matrix, agent loop, structured output, tool approval, streaming, testing, and migration traps. Use when building or upgrading AI SDK code.

• accessibility

  Web accessibility patterns and WCAG 2.1/2.2 compliance. Use to build accessible components, audit a UI for a11y issues, fix screen reader or keyboard problems, check contrast, or add ARIA.

• context-engineering

  How to budget and structure what a coding agent sees. Use when asked to trim, split or write AGENTS.md, CLAUDE.md, rules files or skill descriptions (for example 'CLAUDE.md is too long'), when an agent ignores instructions, forgets rules or drifts in long sessions, and when planning compaction or pr

• debugging

  Systematic debugging by root cause analysis. Use when facing a bug, test failure, flaky behavior, production error, or integration failure. Covers reproduce, hypothesize, fix, prevent, and post-mortems.

• docs-lookup

  Looks up current, version-correct docs before answering a library question from memory. Use when asked for an exact method signature, argument order, config key, CLI flag, or the call for 'our version' or a named version (for example 'in 7.2'). Check the docs or lockfile first even when sure, becaus

• performance

  Profiling-driven performance optimization: caching, query tuning, N+1 fixes, memory, and load testing. Use to diagnose a bottleneck or set performance budgets.

• security

  Application security principles and OWASP Top 10. Use to review code for vulnerabilities, implement authentication or authorization, handle secrets, configure security headers, prevent injection (SQL, XSS, CSRF), or prepare for an audit.

• verification-before-completion

  Gate before saying work is done. Use whenever the user says 'mark it done', 'should work now', 'it's fixed' or 'ready', or asks for a PR summary or handoff note, and before any claim that code, config, docs, a migration or a bugfix is complete. Run the check, read the output, then report with eviden

• finishing-branch

  Finishes a development branch by verifying tests and lint, reviewing the diff, choosing merge, squash, rebase, or stacked PR, writing the PR message, opening the PR, and cleaning up. Use when the user says 'ready to merge', 'finish this branch', or 'open a PR'.

• game-feature-loop

  Add a mechanic, implement a game feature, build a level, make the jump feel better, or prototype a game, one verified step at a time on any engine (Godot, Unity, Unreal, web, others). Detect the engine, plan one mechanic, build, run, prove it from live state, commit, then a separate game-feel pass.

• handoff

  Writes a short handoff note when a session ends mid-task so the next session or person can continue without re-discovery, and resumes from one. Use when asked to hand off, save progress, or continue from a handoff note.

• hooks-and-guardrails

  Adds guardrails to an AI coding agent setup, including deny rules for secrets, sandbox and permission choices, and hooks that block .env reads and gate 'done' on passing tests. Use when hardening an agent, adding hooks, or stopping secret reads.

• pr-feedback

  Works through review comments and failing CI on an open pull request, triaging each comment, fixing with tests, replying on every thread, and looping until CI is green. Use when asked to address review comments, fix CI, or handle PR feedback.

• security-research

  Bug bounty and vulnerability research workflow. Use to find and report a vulnerability, set up a bug bounty workflow, do coordinated disclosure, get a CVE, audit an open-source project you run locally, or hunt variants of a fixed bug. Authorized targets only: an in-scope bounty asset, code you run i

• spec-driven-development

  Turns an approved idea or ticket into spec, plan, and task files saved in the repo, then executes them one task at a time with a progress log. Use to write a spec, plan a feature, break work into tasks, or resume multi-session work.

• tdd

  Red-green-refactor procedure: write the failing test first, show it fail, write the minimum code, refactor on green. Use for any bug fix or new behavior in code that has or could have tests, including 'fix this bug', 'add a function', 'use TDD' and 'write the test first'.

• agent-creator

  Interactive tool to design a sub-agent definition (specialist, role, or team-lead) with the right frontmatter, tools, isolation, memory, and system prompt. Use when asked to create a sub-agent, write an agent file, or scaffold a role agent.

• agentic-rules-writer

  Writes or trims agent instruction files (AGENTS.md, CLAUDE.md, Cursor, Copilot, Codex and more). Use to set up agent rules, or when asked to trim my CLAUDE.md, shorten a rules file that is too long, or clean up AGENTS.md. Keeps only what the model cannot guess. Hard limits become hooks.

• audit-dependencies

  Runs the dependency-auditor agent on a project and returns vulnerabilities, outdated packages, and license findings as JSON. Use for CI, scheduled, or headless runs, for example `claude -p "/agents-virtuoso:audit-dependencies"`.

• brainstorming

  Interactive design exploration before implementation. Use when the user has a vague idea or feature request and no written spec. Asks one focused question at a time, writes a spec, and blocks implementation until approved. Triggers include 'I want to build', 'let's brainstorm', 'help me spec out'.

• communication-style

  Writes messages in the user's own voice, not a generic AI tone. Apply before writing ANY message meant to sound like the user: a GitHub PR or review comment, an issue or thread reply, a Slack or chat DM, an email, or a ticket comment. Loads the user's saved default voice, and when a recipient is nam

• find-test-gaps

  Runs the test-gap-analyzer agent on the files changed in a branch, or on a directory, and returns missing test cases as JSON. Use for CI or headless runs, for example `claude -p "/agents-virtuoso:find-test-gaps origin/main"`.

• plan-migration

  Runs the migration-planner agent on pending database migrations and returns a risk-rated execution plan with rollback paths as JSON. Use for CI or headless runs, for example `claude -p "/agents-virtuoso:plan-migration origin/main"`.

• plugin-creator

  Interactive tool to scaffold a Claude Code plugin with plugin.json, skills, agents, workflows, hooks, MCP servers, and an optional marketplace entry. Use when asked to create a plugin, convert .claude/ configuration into a plugin, or package skills and agents for distribution.

• pr-message-writer

  Writes pull request messages with a technical summary, testing instructions, and verification queries, using customizable templates. Use when writing a PR description, preparing a pull request for review, or documenting changes for teammates.

• report-writer

  Generates a standalone HTML report summarizing changes, findings, debug investigations, or architectural decisions, and opens it in the browser. Use after completing work on a ticket, investigation, or debug session.

• research

  Researches a question across local files, official docs and the web, then reports sourced findings with confidence. Use whenever the user asks to research, compare or choose between tools, find best practice or an industry standard, check what changed recently, or asks what people use, even if they

• review-pr-cold

  Runs the cold-reviewer agent on a diff with no project context and returns the findings as JSON. Use next to review-pr in CI for a second, independent opinion, for example `claude -p "/agents-virtuoso:review-pr-cold origin/main"`.

• review-pr

  Runs the reviewer agent on a pull request or branch diff in an isolated subagent and returns the review as JSON. Use for CI or headless runs, for example `claude -p "/agents-virtuoso:review-pr origin/main"`.

• scan-refactors

  Runs the refactor-scout agent on a directory or on the files changed in a branch and returns code smells and refactoring techniques as JSON. Use for CI, scheduled, or headless runs, for example `claude -p "/agents-virtuoso:scan-refactors src"`.

• skill-creator

  Writes or updates an agent skill (SKILL.md plus references) after researching the topic and verifying every checkable fact. Use when asked to create a skill, write a skill for X, make a new skill, turn this into a skill, or update this skill.

• skill-measure

  Measures whether a skill or reference resource changes a strong model's output enough to earn its context cost, using a fair-fixture A/B ablation and a blind judge. Run by name to audit a skill before keeping, trimming, or removing it.

• stakeholder-update-writer

  Writes a stakeholder Slack or email update message. Use when the user wants to communicate project status, blockers, decisions, or release readiness to stakeholders.

• ticket-writer

  Interactive tool that writes a ticket of the right type (story, subtask, bug, epic, or initiative) with its own structure and quality checks, ready to paste into any tracker. Use when asked to write a ticket, draft a user story, file a bug, or scope an epic.

• ui-review

  Reviews UI code and the rendered page against a pinned, sourced rule set and reports findings with file:line, rule id, severity and a fix. Use when asked to review my UI, audit this page, check my frontend, or check a URL for focus, forms, contrast, motion, touch targets or layout shift.

• verify-acceptance

  Runs the acceptance-verifier agent to map a diff against acceptance criteria and returns a PASS/FAIL/PARTIAL/UNTESTED matrix as JSON. Use for CI or headless runs, for example `claude -p "/agents-virtuoso:verify-acceptance 123 origin/main"`.